A Highly Sensitive, Stretchable and Robust Strain Sensor based on Crack Propagation and Opening

Abstract

Soft and stretchable strain sensors have been attracting significant attention. However, the tradeoff between the sensitivity (gauge factor) and the sensing range has been a major challenge. In this work, we report a soft stretchable strain sensor with an unusual combination of high sensitivity, large sensing range, and high robustness. The sensor is made of silver nanowire network embedded below the surface of an elastomeric matrix (e.g., polydimethylsiloxane). Periodic mechanical cuts are applied to the top surface of the sensor, changing the current flow from uniformly across the sensor to along the conducting path defined by the open cracks. Both experiment and finite element analysis are conducted to study the effect of the cut depth, cut length, and pitch between the cuts. The stretchable strain sensor can be integrated into wearable systems for monitoring of physiological functions and body motions associated with different levels of strains, such as blood pressure and lower back health. Finally, a soft 3D touch sensor that tracks both normal and shear stresses is developed for human-machine interfaces and tactile sensing for robotics.

Graphical Abstract

Introduction

Soft and stretchable strain sensors have been attracting significant interests because of their wide applications in wearable electronics for physiological monitoring and motion tracking^1-12, human machine interfaces^13-16, and soft robotics^17-21. A variety of nanocomposite materials have been explored for use in soft and stretchable sensors, including carbon nanotubes (CNTs)^22-24, graphene^25-27,metal nanoparticles²⁸, magnetic nanoparticles^29-33, and metal nanowires^{10, 34-37}. In stretchable strain sensors, these materials typically exist in the form of three-dimensional percolation network embedded in a polymer matrix. Stretching the composite matrix decreases the percolation density, resulting in decreased electrical conductivity. In parallel, structural design strategies have been used to significantly reduce the effective strain on the conducting materials, leading to better stretchability. A number of mechanically guided structural designs have been reported, such as wrinkles^38-39, serpentines⁴⁰, spirals/helices^41-42, and kirigami^43-44 and auxetic metamaterials⁴⁵.

Recent advancements in wearable technology and soft robotics have led to the development of new stretchable strain sensors with enhanced sensing performances, such as large gauge factor (GF), large strain range, high linearity, better conformability, and robustness³. However, tradeoffs between these performances have been a major challenge for almost all existing strain sensors. For example, a Pt/polyurethane acrylate (PUA) composite resistive sensor with surface crack design has been reported with an ultrahigh GF of 2000, yet a strain range of only 2%⁴⁶, which is low for most applications such as body motion tracking. Another crack-based strain sensor also shows good sensitivity (GF=9400) but with small sensing range (<3%)²⁰. These examples illustrate promising potential of measuring contact resistance change between crack surfaces for strain sensing. However, using well-defined, periodic mechanical cuts in soft materials to control the crack propagation and opening has not been explored. On the other hand, A highly stretchable capacitive strain sensor enabled by wrinkled gold films showed a strain range of 250% but a GF of only 0.9⁴⁷, which limits the sensing capabilities for small strains. Some researchers have also developed sensors with large strain range (450% to 500%) and high sensitivity (GF of 67.7 to 10000)^{38, 48}. But they typically showed a nonlinear response (smaller GF at low strains and large GF at higher strains).

For monitoring of human physiology and motion, the skin strain ranges from less than 1% to over 50%⁴⁹. Typically, one sensor is used to detect subtle strains associated with physiological functions (e.g., blood pulse and respiration), while another for large strains associated with bending of body parts (e.g., knee, elbow, and back). It would be of great interest to develop one strain sensor, in the form of either a skin patch or a textile band, which can capture the full range of strains on human skin and can thus be attached onto different parts of the body⁵⁰. For monitoring certain diseases, it is indeed critical to accurately measure a wide range of strains. For example, in the case of Parkinson’s disease, the symptoms include resting tremor, rigid muscles, bradykinesia and sometimes a combination of above^51-53. The sensors must be sensitive enough for monitoring small tremors while maintaining a large sensing range to measure joint movements. Moreover, soft strain sensors may encounter adverse conditions during operation, such as impacts and over-extension, where the applied strain could substantially exceed the sensing range. Therefore, it is of important relevance to develop a stretchable strain sensor with high sensitivity, large sensing range, and high robustness (i.e., surviving over-strain and repeatability), which remains a significant challenge.

This work reports a soft stretchable strain sensor that can achieve all three goals by combining the use of nanomaterials with a novel structural design strategy. The sensor is made of silver nanowire (AgNW) network embedded below the surface of polydimethylsiloxane (PDMS). Periodic mechanical cuts are applied to the top surface of the AgNW/PDMS composite, changing the current flow from uniformly across the sensor to along the conducting path defined by the open cracks. Under the applied strain, the resistance increases as the crack propagates, but remains constant as the crack reaches the cut length and exhibits tip blunting. Both regions are totally reversible and repeatable. Both experiment and finite element analysis have been conducted to study the effect of the cut depth, cut length, and pitch between the cuts, highlighting excellent tunability of the GF, sensing range, and reversible range of the sensor. The stretchable strain sensor was integrated into two systems for wearable monitoring of blood pressure and lower back health, demonstrating the capabilities for small-strain and large-strain sensing, respectively. Finally, a soft 3D touch sensor that tracks both normal and shear stresses was developed for human-machine interfaces and tactile sensing for robotics.

Results

Fig. 1a shows the fabrication process of the AgNW strain sensor with the surface crack design. The AgNW solution was drop-casted on a sacrificial substrate and then heated to evaporate the solvent. Then liquid PDMS precursor was mixed thoroughly and spin-coated on top of the AgNW network. The AgNW/PDMS composite was cured at 70°C for 1 hour. The cured sample was then peeled off from the substrate with the AgNW network embedded below the surface of the PDMS matrix⁵⁴. Then the composite film was cut from the AgNW side using a mechanical cutter, cutting through the AgNW/PDMS composite into the pure PDMS layer (Fig. 1b). The AgNW/PDMS layer was cut into a zigzag-shaped pattern. Finally, Cu lead wires were attached to the two ends of the pattern by silver epoxy.

Fig. 1. Overview of the soft stretchable strain sensor.

a, Fabrication process of the sensor. b, Electrical model of the sensor during and after cuts. The SEM image shows the cross-section view of one cut surface. c, The undeformed sensor and the sensor under twisting, bending and stretching.

Fig. 1b shows the electrical models of the samples during and after the mechanical cuts. 2R₀ is the initial resistance of the sample. As the razor blade cuts through the AgNW/PDMS composite, AgNWs along the cutting path are broken and pushed in. After retrieval of the blade, the AgNWs on the two cutting surfaces form a physical contact with the contact resistance of R_c. The SEM image shows the cross section view of the cutting surface.

Fig. S1 shows the resistance changes of the AgNW/PDMS composite samples of different AgNW densities with respect to the number of cuts. All samples showed a gradual increase of resistance with the increasing number of cuts. The sample with AgNW density of 0.5 mg/cm⁻¹ yielded a minimum resistance change of only 15.3% after 5 cuts while the sample with AgNW density of 0.1 mg/cm⁻¹ showed a resistance change of 321.9%. This can be explained by the different fracture mechanisms in the two cases. From Fig. S2 a, the low-density sample showed a clean-cut surface with AgNWs along the cutting path broken and sticking out. The depth of the conductive contact surface in this case was ~3 μm. However, for the sample with high-density AgNW network, the AgNW/PDMS composite was deformed and pushed inwards. In the case of 0.5 mg/cm⁻¹, the depth of the conductive contact surface was over 13 μm (Fig. S2 b). This large contact area enabled the resistance change to be as low as 5% after each cut. In the rest of this work, 0.5 mg/cm⁻¹ AgNW density was chosen for all the strain sensors due to their stable contact resistances. The force-displacement curves of the samples with different AgNW densities are shown in Fig. S3. It can be seen that a higher cutting force is required to break through the composite layer with a higher AgNW density. Fig. 1c shows the crack-based strain sensor under twisting, bending, and stretching (30% strain). With increase of the applied strain, the crack gradually propagated, which caused R_c to increase; inversely, with decrease of the applied strain, the crack retracted leading to decrease of R_c. The electromechanical response of the sensor is highly reversible under cyclic loading (to be discussed in Fig. 2).

Fig. 2. Electrical performances of the strain sensor under applied strain.

a, Schematic of the sensor showing the pitch between cuts p, crack length l_c, specimen width w, cut depth d_c and specimen thickness t. b, Resistance change of the sensor corresponding to snapshots in d. c, Sensor response to 1000 cycles of 30% applied strain. The inset figure shows 20 cycles. d, Snapshots of the top surface of the sensor under applied strain with magnified photographs showing the crack opening and tip blunting.

To study the electrical performance of the sensors for different geometrical designs, we defined three major geometrical parameters – ratio of the cut depth to the sample thickness d_c/t, ratio of the pitch between the cracks to the specimen width p/w, and ratio of the cut length to the specimen width l_c/w (Fig. 2a). Fig. 2b shows the resistance change of the crack-based sensor as a function of the applied strain. The resistance change curve can be divided into two regions, a sensing region where the resistance increases linearly with the increasing strain, and a plateau region. The crack propagation/opening process was observed in-situ under an optical microscope, while the resistance was measured simultaneously. 1,000 stretching/unloading cycles (1.2 mm/s) were applied to show the excellent repeatability of the sensor for long term use (Fig. 2c). Fig. S4 shows the stress-strain curve and the resistance change of the sensor under different strain rate (1.2 mm/s, 0.6 mm/s and 0.3 mm/s), which shows good repeatability in all cases.

The optical images in the sensing range (Fig. 2d ①-⑨) show the gradual propagation of a single crack. In this range, the current flow follows the cut guided path. As a result, the resistance of the sensor increases with the crack propagating. However, this resistance change stops when the crack is fully open, i.e., the two crack surfaces lose contact, as shown in the magnified optical image (Fig. 2d ⑩). Fig. S5 also shows the relationship between the applied strain, the crack length, and the relative resistance change.

The sensor can be further stretched without resistance change, as shown in the plateau region in Fig. 2b. In this case, the crack reaches the full cut length. Under the applied strain, the crack further opens with the crack tip blunting but not advancing beyond the cut length. This plateau region is important as it can protect the sensors from failing due to unexpected large strains. Here we define this reversible range by the combination of the sensing region and the plateau region. Beyond the plateau region, the local strain in front of the crack tip would cause irreversible sliding in the AgNW network, leading to overall irreversible resistance change. If the applied strain further increases, the crack tip could propagate beyond the cut length, causing more serious irreversibility in the resistance.

Fig. 3a shows the relative resistance changes of three samples with d_c/t of 1/8, 1/4, and 1/2, with p, w and l_c remaining the same (p/w = 0.2, l_c/w = 0.9). The three cases showed the same GF of 81.5 and the linear sensing region gradually increased with the decreasing cut depth. This is due to the difference in the crack propagation rate for the different cut depths. The shallower the cuts, the larger the applied strain to complete the crack propagation and opening. The plateau region gradually decreased with the decreasing cut depth (Fig. 2d).

Fig. 3. Geometrical design of the strain sensor.

a, Resistance change of the sensors with d_c/t = 1/8, 1/4, and 1/2 with respect to applied strain. Three cross markers with corresponding color represent the reversible range of each case from FEA. b, Resistance change of the sensors with p/w = 0.1, 0.2 and 0.3 with respect to applied strain. c, Resistance change of the sensors with l_c/w = 0.8, 0.85, 0.9 and 0.95 with respect to applied strain.

In Fig. S6, finite element analysis (FEA) (ABAQUS, version 2017) was conducted to simulate the crack propagation and opening behavior of the AgNW/PDMS composite. With the introduction of mechanical cuts, the effective modulus reduces. With larger d_c/t, the effective modulus reduces more. Similarly, with smaller p/w and larger l_c/w, the effective modulus drop of the sensor increases (Fig. S7). The maximum strain in the AgNW/PDMS layer lies ahead of the crack tip during stretching. When it exceeds a critical strain ε_c, we assume that the irreversible sliding in the AgNW network occurs, causing irreversible damage in the AgNW network and hence the irreversible resistance of the sensor. The critical strain on AgNW/PDMS composite was measured to be ε_c = 5% under tension. As shown in Fig. S8, the resistance of the AgNW beyond 5% tensile strain was irreversible. Here we characterize the reversible range by the applied strain when the maximum strain at the crack tip of the AgNW/PDMS layer reaches 5%. FEA simulations found the reversible ranges of 31.4%, 38.3%, and 45.1% for d_c/t of 1/8, 1/4, and 1/2, respectively (marked with cross marks in Fig. 3a). To guarantee the reversibility of the sensor, the applied strain remains below the reversible range, which is much smaller than the breaking strain; above the breaking strain the crack would further advance beyond the initial cut length. To sum up, with different cut depths, a tradeoff exists between the sensing range and the reversible range while having the same GF.

Fig. 3b demonstrates the effect of p/w on the relative resistance change, while d_c/t and l_c/w stayed as 1/8 and 0.9, respectively. With the decrease of p/w from 0.3 to 0.1, the GF increased from 81.5 to 290.1. With a smaller p/w, the pitch between the cracks decreases while the number of the cut paths increases, thus the overall resistance change can increase dramatically. Fig. 3c shows a slight increase of the resistance change with the increasing l_c/w while keeping d_c/t = 1/8 and p/w = 0.3, because the final “neck” of the conducting path after the crack propagation completes is narrower with the increase of l_c/w. These results on the three parameters, d_c/t, p/w and l_c/w, demonstrated excellent tunability of the GF, sensing range, and reversible range of the sensor.

Fig. 4 shows a benchmark comparison between the present work and the reported stretchable strain sensors enabled by different sensing mechanisms, including resistive sensors^{22-23, 26, 28, 35, 46} capacitive sensors^10-11 and piezoelectric sensors⁵⁵. Of note is that the selected strain sensors in the map all showed high linearity and reversibility. In general, a tradeoff exists between the GF and the sensing range for stretchable strain sensors. From this map, it can be seen that the present work yields a large GF = 290.1 with a decent strain sensing range (22%), exceeding the envelop defined by the reported stretchable strain sensors. After integration on stretchable substrates (e.g., athletic tape to be discussed later), the stretchability of the sensor can be further enhanced.

Fig. 4. Performances of the strain sensors compared with the reported stretchable strain sensors.

A coupled electromechanical FEA model was built (COMSOL Multiphysics 5.6) to understand the resistance change of the strain sensor under stretching (Fig. S9 a). The AgNW/PDMS composite was modeled as an isotropic conductive material and the contact resistance between the two crack surfaces was considered to be dependent on the crack opening. When the distance between the two surfaces exceeds a critical value, the contact resistance drops from an initial value to zero. This model can simulate the gradual propagation and opening of the crack. Material parameters used in the simulation such as sheet resistance of the AgNW/PDMS composite and contact resistance were measured (e.g., in the case of NW density of 0.5 mg/cm², sheet resistance of the AgNW/PDMS composite is 0.25 Ω sq⁻¹ and contact resistance per unit length is 1.6 Ω cm⁻¹). By taking the integral of the ECD along the conducting path we can calculate the resistance change with respect to the applied strain. The FEA results agreed well with the experimental results of different p/w ratios (Fig. S9 b). Fig. S9 c shows the simulated electrical current density (ECD) field of the strain sensor before and after stretching. With the increase of the applied strain, the ECD field shows a transition from uniform current flow (across the closed cracks) to current flow along the conducting path defined by the open cracks. This model could predict the performance of the sensors with even more complicated cutting patterns, providing valuable guidance to the sensor design.

To demonstrate the versatile applicability of our strain sensors for monitoring human motions, we first applied the sensors on the wrist to detect the pulse wave, which represents one of the most delicate strain signals on human skin. Fig. 5a shows the setup consisting of a rubber band with a slot filled with the strain sensor. The rubber band was fastened on the wrist to secure the strain sensor against the pulsing area. When the blood pumps through the vein, the two ends of the strain sensor are fixed by the rubber band while the middle is bent and stretched. Then the cracks are opened on the top surface. The red curve in Fig. 5b shows the pulse wave captured from the radial artery on the wrist. Similarly, another strain sensor can be placed on the brachial artery on the arm, recording another pulse wave simultaneously (black curve in Fig. 5b). By measuring the distance between the two pulse areas and taking the average of the time gap between peaks of the two pulse waves, the averaged pulse wave velocity (PWV) can be measured, based on which the blood pressure (BP) can be obtained following⁵⁶

$$\text{BP}=\alpha {\text{PWV}}^2+\beta$$

where $\alpha =0.18\text{kPa}\cdot {\mathrm{s}}^2\cdot {\mathrm{m}}^{-2}$ and $\beta =2.7\text{kPa}$⁵⁶. Taking $\text{PWV}=7.5\text{m/s}$ as measured from Fig. 5b, the BP was calculated to be 12.8 kPa, which equals 96.2 mmHg.

Fig. 5. Demonstrations of the strain sensor for personal health monitoring.

a, The sensor integrated with a wristband for monitoring blood pulse. b, Sensing results of two blood pulse sensors placed on the radial artery on the wrist and the brachial artery on the arm. c, Photographs of the sensor integrated with an athletic tape attached side by side along the spine for monitoring lower back strains. Inset shows the two sensors and the Bluetooth evaluation board for data collection and transmission. d, Sensing results of the two-sensor system detecting different bending motions of the lower back corresponding to photographs in c.

The other demonstration aimed to monitor the large strains on the lower back, which is a critical signal for metabolic syndrome and spine issues^57-58. The sensor was integrated onto an athletic tape by fixing the two ends of the sensor on the tape using a PDMS precursor (Fig. 5c inset). Two sensor/tape setups were attached on the lower back side by side in parallel with the spine. An Arduino Bluetooth board/tape was attached in the middle area of the lower back to collect and transmit the sensing signals to a smart phone. The details of the Bluetooth board design are shown in Fig. S10, where R₀ is the dummy resistor and R₁-R₆ are the six available channels (only two used in this demonstration). The subject started from a sitting-straight posture and leaned forward three times with increasing degree. Then the subject leaned forward while tilting rightward and leftward. Fig. 5c shows the measured strain signals of the two sensor/tape setups. When leaning forward, both sensors responded with resistance increases. While leaning forward and tilted sideways, the resistance of the sensor in the corresponding side remained nearly constant and the one at opposite side increased substantially. The calibration of the sensor/fabric setup is shown in Fig. S11 a. Four fiducial markers were drawn alongside the crack-based sensors on the skin for the purpose of validation. The results measured by the sensor and by the fiducial markers showed an excellent agreement (Fig. S11 b). Of nots is that by integrating the strain sensor with a stretchable athletic tape, the overall stretchability of the sensor/tape setup can be increased on demand according to the need of the applications. In the setup, the strain distributions in the sensor/tape region and the pure tape are different. Due to the larger Young’s modulus, the sensor/tape region has a smaller strain than the pure tape region. By adjusting the modulus ratio, a sensor/tape setup can be easily integrated for wearable applications with large stretchability, e.g., motion tracking on human joints.

Finally, besides wearable personal health monitoring, the senor can be applied for human-machine interfaces and robotics. Fig. 6a shows the schematic diagram of an integrated 3D touch sensor. The two solid AgNW/PDMS composite in the middle form a capacitive pressure sensor that detects normal pressure⁸, while the four surrounding AgNW/PDMS composite ribbons with the surface crack design (CS1, CS2, CS3 and CS4) work as shear stress sensors. The pressure sensor was connected to a capacitance evaluation board and the shear sensors were connected to a multichannel resistance meter (Fig. 6b). Fig. S12 shows the calibration results of the shear sensors. When a shear force is applied in the middle area, one shear stress sensor away from the moving direction is stretched. The cracks in the stretched sensor open, resulting in a resistance increase. Fig. 6c shows excellent decoupled responses when shear forces in four directions were applied sequentially. Fig. 6d shows real-time control of an airplane in a video game with the functions of turning left, turning right, acceleration, and deceleration. This device illustrates a promising potential in virtual reality and soft human-machine interfaces.

Fig. 6. Demonstration of the strain sensor for human-machine interface and tactile sensing.

a, Schematic layout of the 3D touch sensor consisting of a capacitive normal strain sensor and four shear strain sensors. b, Photograph of the 3D touch sensor connected to an evaluation board and laptop. Inset shows the 3D touch sensor. c, Shear strain of the touch sensor when pushing towards different directions using a finger. d, Application in playing an airplane video game. The four sensors correspond to four different functions: turning left, turning down, acceleration, and deceleration, respectively. e, Snapshots showing different stages of grasping a glass of water including grabbing, lifting and dropping. f, Shear strain and normal strain of the tactile sensor during two trials of grasping of the cup of water, showing the grabbing, lifting and dropping.

Fig. 6e shows the sensor applied for tactile sensing. The sensor was integrated on the fingertip of a glove and then used for grasping a glass of water (200 g). Fig. 6f captures the shear and normal strains of two trials with different lifting and dropping speeds. When the glass of water was grabbed but not lifted, the capacitance of the middle sensor increased while the resistance of the shear sensor remained constant. During lifting, the shear strain increased to overcome the gravity of the glass of water. When dropping the water on the table, the shear strain decreased to zero followed by the normal strain decreasing to zero. Note that the shear strain curve of the first trial showed a period of overshot because the fast lifting required extra shear force for acceleration. However, the second trial was 3 times slower in the lifting speed and the shear strain showed no obvious overshot. This application demonstrated the great potential for tactile sensing for robotics.

Conclusions

In summary, this work reported simple, facile fabrication of a versatile soft stretchable strain sensor for wearable applications. We designed and fabricated a AgNW/PDMS composite based strain sensor with simple mechanical cuts into the top surface. Under the applied strain, the resistance increased as the crack propagated (the sensing range), but remained constant as the crack reached the cut length (the reversible range). Under further loading, the local strain in front of the crack tip would cause irreversible sliding in the AgNW network, leading to irreversible resistance change. The effect of the cut depth, cut length, and pitch between the cuts, were studied to optimize the sensor performances. This sensor overcame the limitation of most existing strain sensors and offered unprecedented combination of GF, strain sensing range and robustness (under over-strain and 1,000 repeated loading cycles). A large GF of 290.1 was achieved with a sensing range over 22%. FEA was conducted to validate the electrical performance and predict the mechanical damage, agreeing very well with the experimental results. As demonstrations, the stretchable strain sensor was integrated into several systems for wearable monitoring of blood pressure and lower back health and 3D touch sensing that tracks both compressive and shear stresses simultaneously, illustrating the promising potential for a range of applications including personal human health monitoring, human-machine interfaces, and tactile sensing for robotics.

Methods

Synthesis of AgNW solution.

First, 60 mL of a 0.147 M PVP (MW ~ 40000, Sigma-Aldrich) solution in EG was added to a flask, to which a stir bar was added; the solution was then suspended in an oil bath (temperature 151.5 °) and heated for 1 h under magnetic stirring (150 rpm). Then 200 μL of a 24 M CuCl₂ (CuCl₂۰2H₂O, 99.999+%, Sigma-Aldrich) solution in EG was injected into the PVP solution. The mixture solution was then injected with 60 mL of a 0.094 M AgNO₃ (99+%, Sigma-Aldrich) solution in EG⁵⁹.

Fabrication of the strain sensor.

The AgNW solution was drop-casted on a plasma treated glass slide, which was placed onto a hot plate at 50 °C to evaporate the solvent. After the solvent was evaporated, liquid PDMS (SYLGARD 184, DOW Inc.) with a weight ratio of 10:1 was mixed thoroughly and dropped on top of the AgNW film in a rectangular mold. The AgNW/PDMS composite was cured at 70°C for 1 hour^{10, 60}. The cured sample was then cut from the top surface using a mechanical cutter (silhouette CAMEO). Cu wires were attached to the two outer ends of the sample by silver epoxy (MG Chemicals).

Fabrication of the 3D touch sensor.

First, the as-synthesized AgNWs were drop casted onto a sacrificial substrate. The AgNWs together with the sacrificial substrate were then laser cut into a cross-shaped pattern. The pattern defined the conductive area into five regions: a square region in the middle (one electrode of the capacitive sensor) and four surrounding rectangular regions (for making the crack-based resistive strain sensors). PDMS was poured on top of the five pieces of AgNWs and then baked at 70°C for 1 hour. After curing of PDMS, the sample was flipped over and cut on the designed areas using the mechanical cuter. Another piece of AgNW/PDMS composite was glued on top of the middle square region to compose a capacitive pressure sensor. Then all five parts of the touch sensor were wired with copper wires.

Video game control using the 3D touch sensor.

To transmit the resistance signal, we used an ARDUINO board (nano 33) and built a simple voltage divider circuit with four dummy resistors. Each channel of the touch sensor was connected to a dummy resistor in series. The voltages across dummy resistors and crack-sensors were collected by the ARDUINO board and processed to calculate the real time resistances of the sensors in each direction. An ARDUINO code was built to read the real-time resistance values of each channel and compare to a triggering threshold Rt to decide whether or not to send out a signal. For example, when the 3D touch sensor was pressed and pushed to the right direction, the left crack sensor CS3 reached a resistance increase over 20% which triggered the ARDUINO board to send out a keyboard signal ‘right arrow’. As a result, the airplane in the video game will make a right turn. Similarly, the other three crack-based sensors have been assigned to ‘left arrow’, ‘up arrow’, and ‘down arrow’. In the airplane video game, the four arrow keys represent the functions of turning left, turning right, acceleration and deceleration.

Supplementary Material

Movie 1

Movie 1. Real-time control of an airplane in a video game with the functions of turning left, turning right, acceleration, and deceleration

Movie 2

Movie 2. Robotic sensing of grabbing, lifting, and dropping a glass of water